1 / 17

Darwinian Data Structure Selection

Darwinian Data Structure Selection. ARTEMIS. Michail Basios Lingbo Li Fan Wu Leslie Kanthan Earl T. Barr. Donald Knuth’s Advice. Darwinian Data Structure Selection Speaker: Michail Basios. Focus Is on User Requirements.

fthomas
Download Presentation

Darwinian Data Structure Selection

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Darwinian Data Structure Selection ARTEMIS Michail Basios Lingbo Li Fan Wu Leslie Kanthan Earl T. Barr

  2. Donald Knuth’s Advice Darwinian Data Structure Selection Speaker: Michail Basios

  3. Focus Is on User Requirements Darwinian Data Structure Selection Speaker: Michail Basios

  4. It’s Not Just Pre-optimization Darwinian Data Structure Selection Speaker: Michail Basios

  5. Optimisation Can Be Beneficial Darwinian Data Structure Selection Speaker: Michail Basios

  6. Data Structure Selection/Optimisation Process Hmmm, maybe just use defaults that work in most cases. Darwinian Data Structure Selection Speaker: Michail Basios

  7. Darwinian Data Structure Selection ARTEMIS Corpus: 43 projects 4.8% median execution time improvement 10.2% median less memory consumption 5.1% median less CPU usage • 13% execution time improvement • 9% less memory consumption • 4% less CPU usage - 23.5% less memory consumption (without affecting other measures) - 16.5% less memory consumption (without affecting other measures) A framework for multi-objective code optimisation Speaker: Michail Basios

  8. Motivating Example Data Structure Implementation Size Parameter A function from the Splitter class of Guava Library Darwinian Data Structure Selection Speaker: Michail Basios

  9. Artemis searches a program’s space of Darwinian Data Structure holes for optimal implementations and configurations. Darwinian Data Structure Selection Speaker: Michail Basios

  10. Architecture of Artemis Source to Source Transformation Darwinian Data Structure Selection Speaker: Michail Basios

  11. What is the improvement that ARTEMIS provides for each measure? Best improvements per measure Popular GitHub projects Darwinian Data Structure Selection Speaker: Michail Basios

  12. Cost of Artemis Average optimisation process -> 3.05 hours Azure cost -> £ 0.41 per hour ≪ Software Engineer Cost Artemis cost ARTEMIS Darwinian Data Structure Selection Speaker: Michail Basios

  13. Artemis as a Web Service Artemis Web-Based Dashboard Darwinian Data Structure Selection Speaker: Michail Basios

  14. Artemis as a Command Line Tool Darwinian Data Structure Selection Speaker: Michail Basios

  15. What makes Artemis cool? • Easy to use as it automatically identifies and tunes the data structures and corresponding parameters of a program • Easy adaptable (String Buffer vs String Builder) • Multiple languages (Java, C++, Scala, etc) • Multi Objective optimisation on programs used by thousands of users • It can scale on the cloud and provide fast optimisations • Potential to be a very nice cloud service connected with Git Darwinian Data Structure Selection Speaker: Michail Basios

  16. Darwinian Data Structure Selection Speaker: Michail Basios

  17. Darwinian Data Structure Selection Darwinian Data Structure Selection Speaker: Michail Basios

More Related